This is an open access article distributed under the terms of the Creative Commons
Attribution License (Between 2000 and 2017, the percentage of US citizens who perceived members of Congress of the United States to be dishonest increased from 21% to 60% (Gallup, 2014). Similarly, the percentage of citizens in Europe who trust their respective national political parliament dropped from 38% to 28% between 2004 and 2014 (European Commission, 2014). Both findings are in line with the general observation that political trust has been declining in Western societies (e.g., Dalton, 2004; Hetherington, 1998). A similar picture can be found in Germany where our research has been conducted. In 2017 only 14% of all respondents reported that they trust politicians (GfK, 2018). In addition, this negative trend is also reflected in the subjective perceptions of German citizens: In 2018 65% believed that politicians are less likely to tell the truth compared to 30 years ago (Ipsos, 2018).
An increasing negativity bias in media reports on politicians has been blamed for this decline (Cappella & Jamieson, 1997; Kleinnijenhuis, van Hoof, & Oegema, 2006; Lengauer, Esser, & Berganza, 2011; Moy & Pfau, 2000). For example, the mass media report about politicians being corrupt (Malkin, 2016; Trex, 2008) or incompetent (Brand, 2017; Riggins, 2017) on a regular basis. This raises the question: To which degree do negative media reports on the behavior of specific politicians influence the perceived trustworthiness of that specific politician, of politicians in general, and of the political system?
There is some evidence indicating that media reports on the wrongdoing of single politicians can impair political trust. For example, knowing that a particular politician was involved in a norm violation reduced his or her perceived integrity (Hendry, Jackson, & Mondak, 2009). In addition, increasing the salience of a politician’s past violation of law has been found to reduce citizens’ trust in politicians in general (e.g., Bless, Igou, Schwarz, & Wänke, 2000; Bowler & Karp, 2004), in political institutions, and the political system (Bowler & Karp, 2004). Other findings indirectly speak in favor of this effect. For example, Welch and Hibbing (Welch & Hibbing, 1997) found that citizens’ electoral support for a politician’s party decreased after this politician was proven to be corrupt. However, previous research has not been conclusive. First, some studies have investigated the potential consequences of media reports on morally questionable, but not illegal behavior (e.g., Funk, 1996; Mölders, Van Quaquebeke, & Paladino, 2017). There is reason to believe that these reactions might differ, based on the moral convictions of the perceiver and that they are generally not as strong as reactions to violations of existing law. Second, most studies have only presented norm violations that occurred at some time point in the past; yet, research indicates that reactions to norm violations change over time (Mitchell, 2014). Thus, short-term consequences of such media reports might be systematically underestimated. It needs to be mentioned, however, that we exclusively investigated the effects of media reports about male politicians. Third, studies that have focused on citizens’ immediate reactions to the wrongdoing of a politician (Halmburger, Rothmund, Schulte, & Baumert, 2012; Hendry et al., 2009; Kepplinger, Geis, & Siebert, 2012) have been limited to specific cases and did not take into account the idea that citizens’ reactions might vary by contextual factors such as the type of norm violation (Carlson, Ganiel, & Hyde, 2000; Funk, 1996). Fourth, previous studies have focused exclusively on how such violations affected the public’s perceptions of either (a) the politicians in question (e.g., Funk, 1996; Hendry et al., 2009; Mitchell, 2014) or (b) politicians in general and the political system (e.g., Bless et al., 2000; Bowler & Karp, 2004; Halmburger et al., 2012; Maier, 2011). Thus, little is known about the underlying psychological processes that are involved when laypersons draw inferences from a single politician’s behavior to all politicians in general or the political system. Although such generalizations are highly plausible, only few papers have discussed the psychological processes that might underlie these effects (e.g., Schwarz & Bless, 1992).
The present research was aimed at overcoming these limitations. First, we focused exclusively on reports about behaviors of individual politicians that reflect violations of existing law and not just morally questionable behavior. Second, we investigated the immediate reactions to new media reports about a single politician’s wrongdoing. Third, we used two different kinds of transgressions, namely financial fraud and exchange of child pornography. By doing so, we could test the robustness of the effects. Fourth, we systematically investigated the impact on three kinds of political trust, namely trust in the specific politicians, trust in politicians in general, and trust in the political system. Complementing previous research from U.S. samples that rather focused on trust of specific candidates and research from Germany that more strongly focused on global levels of trust, we conducted a series of studies in Germany that considered the evaluation of a specific politician, all politicians, and the political system of Germany.
We draw from theoretical accounts on trait inferences (Uleman, Newman, & Moskowitz, 1996) in order to hypothesize how information about a politician’s unlawful behavior can feed into the evaluation of his or her trustworthiness. Based on reasoning in social categorization research (Crisp & Maitner, 2011; Tajfel, Billig, Bundy, & Flament, 1971) and the inclusion-exclusion model (Bless & Schwarz, 2010), we argue how reports about a politician’s transgression can translate into the perceived untrustworthiness of politicians in general and the political system.
Trait Inferences: How Reported Law Violations Influence the Trustworthiness of a Politician
In literature on media communication, the influence of media reports on politicians is often explained by processes in which citizens draw inferences from media information to the judgment of a politician (e.g., Entman, 1991; Kepplinger et al., 2012). In line with the psychological understanding on trait inferences (e.g., Clifford, 2014; Uleman et al., 1996), we posit that recipients use media reports about a politician’s behavior to form judgments about his or her character (Cundy, 1994). Empirical research suggests that information about a person’s wrongful behavior (e.g., telling a lie) is typically ascribed to the person’s stable characteristics rather than to specific aspects of the situation (Skowronski & Carlston, 1989). But which character dimension do citizens take into account when they evaluate the trustworthiness of a politician? According to the integrative model of trust (Mayer, Davis, & Schoorman, 1995), trait inferences about integrity, benevolence, and competence should feed into the perceived trustworthiness of a politician (Funk, 1999; Kinder, 1986). Ascribing integrity to a politician is commonly associated with viewing this person as moral, honest, just, and reliable. A benevolent politician is seen as kind, loyal, caring, supportive, and generally responsive to citizens’ interests. A competent politician is perceived as having ability, expertise, knowledge, and problem-solving skills.
Results of a meta-analysis (Colquitt, Scott, & LePine, 2007) indicated that integrity, benevolence, and competence represent empirically separable, albeit correlated facets of trustworthiness. In more detail, empirical results suggest that the relation between these three dimensions of trustworthiness and trust-related behaviors are all mediated by the interactional trust between the trustor and the trustee. In addition, direct relations between these three dimensions with trust-related behavior (e.g., counterproductive behavior, which was positively related to competence, but negatively to benevolence) suggest unique patterns of benevolence, integrity, and competence. This finding is in line with the theoretical assumption that all three dimensions contribute independently to the perceived trustworthiness of a politician. Neither will people trust a politician who acts in line with moral standards and citizens’ interests, but repeatedly makes rationally wrong decisions, nor will they trust a politician with excellent problem-solving skills who invests all efforts for his or her own good. The few empirical studies that conceptualized trustworthiness of politicians as a multidimensional construct also suggest that media reports can have unique impacts on one, but not another dimension of trustworthiness (e.g., Funk, 1999; Hendry et al., 2009). All three dimensions have been found to be important underpinnings of trust in and behavior toward the respective politician in research in the U.S. and Germany (e.g., Funk, 1999; Halmburger, Rothmund, Baumert, & Maier, 2019; Hayes, 2010; Mölders et al., 2017). Also, these findings suggest differences in the relative importance of these dimensions, depending on the specific context or candidate. Thus, on the basis of these findings and theoretical considerations, it seems reasonable to investigate trustworthiness as a multidimensional rather than a global concept. More precisely, we aim to investigate whether and how people draw inferences from a politician’s law violation to his or her benevolence, integrity, and competence. Based on previous research and psychological theory, three different expectations seem plausible.
First, a politician’s law violations might exclusively reduce this politician’s perceived integrity, but not his or her perceived benevolence and competence. In line with previous research (e.g., Clifford, 2014; Uleman et al., 1996), citizens should generally be inclined to draw inferences from media reports about a politician’s law violation to those parts of the politician’s character that are directly related to his or her moral integrity. Thus, it might be exclusively the perceived integrity of the politician that is affected, as this dimension represents the expectation that this politician is moral, honest, just, and reliable (e.g., Funk, 1999; Kinder, 1986). Providing some support for this idea, previous research found that information about political norm violations decreased the perceived integrity, but not the competence of the politicians in question (Hendry et al., 2009). Benevolence was not assessed in this research.
Second, political law violations might reduce perceived integrity and benevolence but not perceived competence. This prediction is based on the notion that people’s perceptions of a politician’s personality differ on two general dimensions, namely warmth and competence (Caprara, Barbaranelli, & Zimbardo, 1997; Wojciszke & Klusek, 1996). According to social perception theory (Fiske, Cuddy, & Glick, 2007), the dimensions of integrity and benevolence can be collapsed into one without excessive loss of information. Thus, a politician’s perceived integrity and benevolence should be strongly related and might be equally affected by information about immoral behavior. Supporting this claim, a politician’s disrespectful behavior was found to decrease his or her perceived warmth, but not competence (Mölders et al., 2017).
Finally, media reports about a politician’s law violation might affect all three dimensions of trustworthiness. Under the label of the halo effect, research has revealed that people have a tendency to make inferences from often irrelevant information about a person’s unknown characteristics (Forgas, 2011; Kaplan, 1978; Nisbett & Wilson, 1977; Sigall & Ostrove, 1975; Thorndike, 1920). More specifically, characteristics generate a halo that touches and therefore influences other characteristics that are not under question. The occurrence of a halo effect has repeatedly been explained by people’s tendency to strive for a cohesive and consistent way to process information and form judgments. In addition, a more rapid, automatic, and constructive processing of information could promote this effect (Forgas, 2011). It can also be assumed that a halo effect is stronger on characteristics that are closely related to those the person received information about than unrelated characteristics. This assumption has already been discussed under the term true halo effect (Murphy, Jako, & Anhalt, 1993). In our example, negative media information about a specific politician could not only influence the perceived integrity of this politician, but generalize to other characteristics. Thus, a halo effect will appear when all three dimensions of the perceived trustworthiness of the politician in question decrease after the perceiver learns that the politician has violated existing law.
Addressing this question is a first step toward understanding the extent to which media reports about a politician’s violation of law are generalized across the three trustworthiness dimensions. In a next step, we take a closer look at information processing in order to understand whether and how citizens potentially generalize from a single politician’s trustworthiness to the trustworthiness of politicians in general and of the political system.
Assimilation: Generalizing From One Politician to All Politicians and the Political System
We conceptualize the perceived trustworthiness of politicians in general as an expectation that politicians as a social group are morally integer, benevolent, and competent (Baumert, Halmburger, Rothmund, & Schemer, 2017; Halmburger, Rothmund, et al., 2019; Schoorman, Mayer, & Davis, 2007). Empirical findings provided evidence for this structural equivalence of trustworthiness in specific politicians and politicians in general (Halmburger, Rothmund, et al., 2019). In a similar vein, the perceived trustworthiness of a political system can be defined as an individual’s expectation that the political system is generally fair, responsive to society’s needs, and efficient (Miller & Listhaug, 1990; Mishler & Rose, 2001; Schiffman, Thelen, & Sherman, 2010). Although inter-individual differences in such generalized expectations yield some degree of stability across time and situations, they are not invariant, but can be influenced, for example by new information (Baumert et al., 2017). In line with theories from personality psychology and political science, trust-related experiences with individuals shape generalized trustworthiness expectations (Mishler & Rose, 2001; Rotter, 1971). As already mentioned, there is empirical evidence suggesting that negative information about a specific politician can affect citizens’ perceptions of trustworthiness of politicians in general (Bless et al., 2000; Régner & Le Floch, 2005; Schwarz & Bless, 1992) and the political system (Anderson & Tverdova, 2003; Bowler & Karp, 2004; Halmburger et al., 2012). But how is new information about a single politician’s behavior integrated into the perceived trustworthiness of politicians and of the political system?
According to social categorization theory (Crisp & Maitner, 2011; Tajfel et al., 1971), receiving information about a specific target (i.e., a politician) should activate the respective social category (i.e., all politicians, the party, and the political system) and expectations regarding this social group or system, including the perceived trustworthiness of prototypical exemplars of this category (Crisp & Maitner, 2011; Tajfel et al., 1971). Conversely, there is also evidence that generalized expectations regarding social categories can be influenced by new information about specific targets. For example, the inclusion-exclusion model (Bless & Schwarz, 2010) posits that the evaluation of a social category (i.e., politicians) that is accessible at the time of this evaluation is influenced by information about specific targets. The evaluation of a social category can be influenced by either the inclusion or the exclusion of information, represented by two underlying processes: assimilation or contrasting. On a general level, assimilation effects reflect the process of adjustment of a given social category to the valence of some piece of information (Bless & Schwarz, 2010). This effect can work for both positive and negative information, either by fostering a more positive or by fostering a more negative evaluation of the social category. According to the inclusion-exclusion model, it can be posited that generalized expectations regarding politicians or the political system can be assimilated to information about a specific politician’s behavior. For example, information about a politician’s law violation could be integrated into the expectations regarding the social category and affect the perceived trustworthiness of politicians in general or even the political system. Alternatively, contrast effects describe a process in which the valence of information and the evaluation of the category have a negative (inverse) relationship (Bless & Schwarz, 2010). As a result, the specific target and the social category or its representatives build two distinct categories that will be compared. The changed evaluation of the social category is caused by this comparison. In other words, negative information about a specific politician can be excluded from the evaluation of politicians in general and the political system, resulting in an even more positive evaluation of all politicians or the political system than without this information.
The inclusion-exclusion model also includes ideas on boundary conditions that decide whether assimilation or contrast effects occur (Bless & Schwarz, 2010). First, for assimilation effects to occur, the accessible information must be perceived as relevant for the evaluation of the category. When people become aware that the information is irrelevant or comes to mind for the wrong reason, however, contrast effects will be more likely. Second, assimilation effects occur more often when the piece of information is somehow representative of the social category. This is when the social category is perceived as rather homogenous, and the target of the information as a subordinate representative of the superordinate category (e.g., one politician out of all politicians). In addition, the perceived representativeness of the target influences the assimilation of the social category to this information.
In general, empirical findings that aimed at testing these theoretical considerations were interpreted as supporting these claims (Bless et al., 2000; Régner & Le Floch, 2005; Schwarz & Bless, 1992). In most of these studies, participants were confronted with names of politicians who were or were not associated with political scandals in the past (scandal activation). Then, participants’ judgments on either politicians in general or the specific other politicians, but not the scandalized politicians, were assessed. In line with the predictions of the inclusion-exclusion model (Bless & Schwarz, 2010), politicians in general were more negatively perceived after scandal activation than without (assimilation effect). However, distinct other politicians were more positively evaluated after scandal activation than without (contrast effect). Empirical results also showed that the more such information is new and extreme (Bless et al., 2000), the more likely and strongly it will influence the perception of the social category. Two points have to be considered critically with regard to these studies. Frist, some of them seem to be highly underpowered considering the complexity of their designs. Second, it remained unclear whether the activation (strength of association) worked for all participants the same way and with the same intensity. According to the empirical findings of Bowler and Karp (Bowler & Karp, 2004), however, the impact of the immoral behavior of politicians on the perceived trustworthiness of the political system highly depends on how much the participants remember this incident. Other studies have overcome these limitations. In these studies, participants were directly confronted with a (fictitious) media report about specific politicians (e.g., Halmburger et al., 2012). Also in these studies, a negative effect on the perceived trustworthiness of politicians in general and the system has been found.
Most important, however, these results do not tell us what happens if participants are allowed to judge the politician in question and the social categories this politician belongs to. Most studies that investigated the theoretical claims of the exclusion-inclusion model (e.g., Bless et al., 2000; Régner & Le Floch, 2005; Schwarz & Bless, 1992) do not discuss how the new information affects the target and how this is related to the decrease of trustworthiness in politicians in general or the political system. In consequence, none of these studies assessed the perceived trustworthiness of the specific politician responsible for the law violation. On the basis of these results, it remains unclear whether the impact of the negative information of the social category (politician in general or the political system) is caused by assimilation effects that occur independently of the evaluation of the specific politician in question or are driven by the negative evaluation. In the first case, we would assume independent direct effects of the law violation on the evaluation of the specific politician and the social category. In the latter case, we would assume that the effect of the law violation on the social category is mediated by the perception of the specific politician.
To replicate previous findings, we tested whether there are effects on the perceived trustworthiness of all politicians and the political system, when participants are confronted with media reports on one specific politician. Moreover, we aimed to test whether this potential effect is mediated by the perceived lack of trustworthiness of the politician in question. We believe that the boundary conditions for an assimilation effect are more likely met than for a contrast effect: information on a specific politician should be relevant for all politicians and all politicians should be seen as the subordinate category of a specific politician rather than a distinct category. In line with this claim, we hypothesized that the more citizens attributed a politician’s law violation to a lack of his or her trustworthiness the more this perception should affect and impair their perceived trust in politicians in general and in the political system.
Present Research
The first aim of the present research was to investigate how laypersons draw inferences about a politician’s trustworthiness from media reports about his or her violation of law. We employed an experimental design to provide a rigorous test of the short-term effects of reading an ostensible news report about a politician’s law violation on the evaluation of his or her integrity, benevolence, and competence. By doing so, we aimed to understand better whether reports about violations of law impair a politician’s trustworthiness in a global way or rather on specific facets. On the basis of theoretical approaches on trait inferences (e.g., Clifford, 2014; Uleman et al., 1996), person perception (e.g., Fiske et al., 2007), and the halo effect (e.g., Forgas, 2011), three alternative predictions were plausible.
Hypothesis 1a: A political law violation will impair the perceived trustworthiness of the politician in question only with regard to his integrity (but not benevolence or competence).
Hypothesis 1b: A political law violation will impair the perceived trustworthiness of the politician in question with regard to his integrity and benevolence (but not competence).
Hypothesis 1c: A political law violation will impair the perceived trustworthiness of the politician in question with regard to his perceived integrity, benevolence, and competence.
The second aim of the present research was to investigate whether and how information about a politician’s law violation is generalized to all politicians and the political system (Halmburger et al., 2012; Régner & Le Floch, 2005). In extension to the inclusion-exclusion model (Bless & Schwarz, 2010), we hypothesized that the perceived untrustworthiness of a single politician can mediate the impact of a media report about this politician’s wrongdoing on the perceived trustworthiness of politicians in general and the political system.
Hypothesis 2: The impact of a political law violation will generalize to the perceived trustworthiness of politicians in general mediated by the perceived trustworthiness of the politician in question.
Hypothesis 3: The impact of a political law violation will generalize to the perceived trustworthiness of the political system mediated by the perceived trustworthiness of the politician in question.
The present research involved four studies. Note that all of the studies were conducted as part of an interdisciplinary research network in order to investigate various research questions, including, for example, the impact of law violations on trustworthiness, emotions, and retribution tendencies or how reactions towards such law violation are moderated by stable personality dispositions. In order to be transparent about our research, we included all materials and measures of the four studies to the Supplementary Material that has been published on osf.io/3gquv (Halmburger, Baumert, & Rothmund, 2019). The present article covers the common ground of these studies, by focusing on the impact of law violations on the different levels of perceived trustworthiness. In each study, we described to the participants a fictitious politician’s law violation in the experimental group versus a behavior that involved no law violation in the control group. As previous studies have suggested that the impact of a politician’s norm violation differs by the domain of the relevant norm (Carlson et al., 2000; Funk, 1996), we used media reports of two different types of law violations (i.e., child pornography or a financial fraud).
All four studies used fictitious, but realistic cases, and were conducted in Germany. Germany is a democratic, federal parliamentary republic. The parliament of Germany (Bundestag) consists of representatives from multiple political parties who are elected for four years. In addition, all sixteen federal states and city states of Germany have their own parliaments (Landtag) with legislative power and competence over the respective state. The city council is constituted by the mayor and the members of the city council (councillors) in cities and communities. In our research, we also varied the politician’s office (member of the Bundestag, member of a federal parliament, city councillor) across our studies. An overview of the content of the provided articles for the experimental manipulation, the respective offices of the involved politicians as well as the settings in which the studies have been conducted can be found in Table 1.
Table 1
Demographic Data for Studies 1 to 4
| Characteristic | Study 1 | Study 2 | Study 3 | Study 4 |
|---|---|---|---|---|
| Setting | ||||
| Online | Laboratory | Online | Laboratory | |
| Experimental condition | ||||
| Law violation | Child porn | Tax fraud | Tax fraud | Child porn |
| Control | School renovation | Holiday | Visiting Bundestag | School renovation |
| Politician’s office | ||||
| City councillor | Member of federal parliament | Member of Bundestag | City councillor | |
| Participants | ||||
| N | 421 | 268 | 139 | 122 |
| Sex | ||||
| Female | 45.5% | 79.5% | 46.0% | 67.2% |
| Age | ||||
| M (SD) | 28.46 (12.06) | 22.53 (3.11) | 47.59 (13.55) | 22.47 (2.34) |
| range | 18-87 | 18-47 | 20-74 | 19-33 |
| Education | ||||
| No | 2.9% | 0.7% | ||
| Secondary | 8.8% | 44.6% | ||
| A levels | 54.2% | 100% | 23.7% | 100% |
| University | 34.2% | 28.8% | ||
| Nationality | ||||
| German | 98.1% | NA | NA | 96.7% |
| Non-German | 1.7% | NA | NA | 3.3% |
| Native language | ||||
| German | NA | 92.2% | 93.5% | NA |
| Other | NA | 5.2% | 6.5% | NA |
In Studies 1, 3, and 4, we tested whether those reports affected the trustworthiness of the respective politician (integrity, benevolence, competence; Hypotheses 1a-c). In Study 2, we tested whether the information about a law violation had an impact on the perceived trustworthiness of politicians in general and the political system. Finally, we investigated whether effects on the trustworthiness of politicians in general (Study 3, Hypothesis 2) and of the political system (Study 4, Hypothesis 3) were mediated by the perceived trustworthiness of the politician in question.
Study 1
In Study 1, we investigated whether a law violation by an individual politician would affect his perceived integrity (Hypothesis 1a), integrity and benevolence (warmth dimension; Hypothesis 1b), or all dimensions of perceived trustworthiness (halo-effect; Hypothesis 1c).
Method
Sample
Participants were recruited via different mailing lists of German universities, but also via different social networks in order to achieve a heterogeneous sample. Out of the 532 persons who started the survey, 433 completed it. Data of two participants had to be excluded because their German language skills were not sufficient. These participants stated that they had poor German language skills (0 on a scale from 0 = poor to 3 = very well). Nine participants were excluded because they were under 18 years of age. The demographic data of the remaining N = 421 participants can be found in Table 1. We used G*Power to conduct power analyses for all studies (Faul, Erdfelder, Lang, & Buchner, 2007). For the conducted ANOVA with manipulation as a first factor, repeated measurement on the second factor, the interaction of the between and within-factor, are based on average correlation of r = .55 for the repeated measures. Post hoc power analyses indicate a power of 1.00 for small (f = 0.10) and medium effects (f = 0.25).
Procedure
First, participants provided demographic information (age, sex, education, occupation, German language skills). They were randomly assigned to read one of two versions of a bogus newspaper article reporting on the behavior of a fictitious German politician. Then, among other items, participants were asked to answer items that assessed the perceived trustworthiness of the politician in question (these and all additional variables can be found in Table 1 of the Electronic Supplementary Material). Finally, participants answered manipulation check items before being thanked and debriefed.
Design and Measures
In this study and all following studies, a one-factorial between-subject design (law violation absent vs. present) was implemented. All items were answered on 6-point rating scales ranging from 0 (totally disagree) to 5 (totally agree).
Bogus newspaper reports
Participants read either a fictitious media report on a politician’s violation of the law (n = 214) or behavior not representing a law violation (n = 207; for both bogus newspaper reports, see Figure 1 and 2 of the Electronic Supplementary Material). In the law violation condition, the participants read that a city councillor was accused of having possessed and exchanged child pornography in several cases. The police had searched the councillor’s office at the town hall. The accused politician had not issued any statement on the accusations. In the control condition, a city council member was entrusted with organizing the invitation to tender and supervise the project of a primary school refurbishment. The mentioned city councillor successfully approved money for this renovation. Layout, reported names, and places presented in the articles were identical. No information about the politician’s party affiliation was given.
Perceived trustworthiness of the politician in question
In line with previous work (Mayer et al., 1995; Mayer & Davis, 1999), we adapted an eight-item scale to assess all three dimensions of perceived trustworthiness, namely integrity (two items: “I don’t believe that the politician keeps his promises,” and “I perceive the politician as little sincere”, both recoded; α = .58, r = .41, p < .001), benevolence (two items: “I believe that the politician aims to achieve the best for the citizens” and “The politician mainly follows own interests”, recoded; α = .61, r = .44, p < .001), and competence (four items: “I believe that the politician is able to solve political problems”, “I am willing to entrust the politician with important tasks”, “I think this politician is competent”, and “I count on the politician being able to act in line with my interests”; α = .87). The results of a confirmatory factor analysis provided evidence that these three dimensions are empirically separable (Mayer & Davis, 1999).
Manipulation check
Participants evaluated whether the politician’s behavior was legitimate (four items, “I view the politician’s behavior as not problematic”, “The politician’s behavior was legitimate”, “The politician’s behavior was proper”, and “The politician’s behavior was prosocial”; α = .96).
Results and Discussion
Manipulation Check
Participants in the law violation condition evaluated the politician’s behavior as significantly less legitimate (M = 0.65; SD = 1.02), compared with participants in the control condition (M = 3.26; SD = 1.09), t(419) = 25.38, p < .001, d = 2.47, 95% CI [2.22; 2.73]. Besides the relative difference between both conditions, we tested the means against the midpoint of the scale (i.e., 2.5) for both groups separately. Participants in the law violation condition clearly perceived the behavior as illegitimate, t(213) = -26.61, p < .001, d = 1.81. Participants in the control condition perceived the behavior as legitimate or prosocial, t(206) = 10.02, p < .001, d = 0.70.
Perceived Trustworthiness of the Politician
Means, standard deviations, and correlations for all variables are displayed in Table 2.
Table 2
Means, Standard Deviations, and Correlations in Study 1 to 4
| Trust dimension | Study 1
|
Study 2
|
Study 3
|
Study 4
|
|||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | 2 | 3 | 4 | 5 | 1 | 2 | 3 | 4 | 1 | 2 | 3 | 5 | |
| 1. Integrity | |||||||||||||
| 2. Benevolence | .53*** | .87*** | .66*** | ||||||||||
| 3. Competence | .43*** | .68*** | .70*** | .64*** | .66*** | .61*** | |||||||
| 4. Politicians | .35*** | .39*** | .46*** | ||||||||||
| 5. System | .64*** | .23* | .40*** | .24** | |||||||||
| M | 2.51 | 2.43 | 2.42 | 2.01 | 2.98 | 1.43 | 1.43 | 2.04 | 1.98 | 2.23 | 2.43 | 2.58 | 2.68 |
| SD | 1.25 | 1.16 | 1.07 | 0.90 | 1.13 | 1.34 | 1.36 | 1.31 | 1.12 | 1.51 | 1.19 | 1.10 | 0.84 |
Note. Response options ranged from 0 (totally disagree) to 5 (totally agree); integrity = specific politician’s integrity; benevolence = specific politician’s benevolence; competence = specific politician’s competence; politicians = trustworthiness of politicians in general; system = trustworthiness of the political system.
*p < .05. **p < .01. ***p < .001.
To test how the report of a political law violation affected the three dimensions of perceived trustworthiness, we calculated a 2 (experimental condition: law violation / control condition) x 3 (dimensions of the perceived trustworthiness of the politician: integrity / benevolence / competence) ANOVA with repeated measures on the second factor. By this, we are able to test in a more straightforward way whether potential effects of the law violation on the three dimensions significantly differ between each other.i Mauchly’s test indicated a violation of sphericity for the repeated measures, χ2(2, 419) = 56.19, p < .001. Therefore, the degrees of freedom were corrected by applying Greenhouse-Geisser estimates.
Results showed a significant main effect of experimental condition, F(1, 420) = 64.78, p < .001, η2p = .13, 95% CI [.08; .19], and a significant interaction of experimental condition and the dimensions of perceived trustworthiness, F(1.78, 744.37) = 12.08, p < .001, η2p = .03, 95% CI [.01; .05]. The main effect for the repeated measures factor was not significant, F(1.78, 744.37) = 1.98, p = .14, η2p = .01, 95% CI [.00; .02]. Table 3 displays means and standard deviations, separately for experimental conditions, as well as results of post-hoc comparisons. In support of a halo effect (Hypothesis 1c), participants who read about a city councillor accused of possessing child pornography perceived this politician as having less integrity, less benevolence, and less competence compared with participants in the control condition.
Table 3
Means and Standard Deviations for Trustworthiness Evaluations Presented Separately for Experimental Conditions in Studies 1 to 4
| Trust dimension | Control condition
|
Law violation condition
|
95% CI(d) | |||||
|---|---|---|---|---|---|---|---|---|
| M | SD | M | SD | t | p | d | ||
| Study 1, N = 421 | ||||||||
| Integrity | 3.02 | 1.14 | 2.02 | 1.16 | 8.99 | < .001 | 0.87 | 0.67; 1.07 |
| Benevolence | 2.69 | 1.11 | 2.18 | 1.17 | 4.58 | < .001 | 0.45 | 0.25; 0.64 |
| Competence | 2.73 | 0.97 | 2.12 | 1.07 | 6.18 | < .001 | 0.60 | 0.40; 0.79 |
| Study 2, N = 268 | ||||||||
| Politicians | 2.38 | 0.95 | 1.75 | 0.77 | 5.71 | < .001 | 0.74 | 0.49; 0.99 |
| System | 3.31 | 1.01 | 2.74 | 1.16 | 4.14 | < .001 | 0.52 | 0.27; 0.77 |
| Study 3, N = 139 | ||||||||
| Integrity | 2.56 | 1.18 | 0.71 | 0.86 | 9.93 | < .001 | 1.86 | 1.45; 2.26 |
| Benevolence | 2.43 | 1.36 | 0.79 | 0.92 | 7.81 | < .001 | 1.48 | 1.09; 1.86 |
| Competence | 2.69 | 1.30 | 1.63 | 1.15 | 5.02 | < .001 | 0.88 | 0.52; 1.23 |
| Politicians | 2.15 | 1.11 | 1.87 | 1.11 | 1.46 | .15 | 0.25 | -0.09; 0.56 |
| Study 4, N = 122 | ||||||||
| Integrity | 3.38 | 0.91 | 1.19 | 1.16 | 11.50 | < .001 | 2.09 | 1.65; 2.53 |
| Benevolence | 3.13 | 0.92 | 1.81 | 1.06 | 7.32 | < .001 | 1.33 | 0.93; 1.72 |
| Competence | 3.18 | .075 | 2.04 | 1.09 | 6.67 | < .001 | 1.21 | 0.82; 1.59 |
| System | 2.78 | 0.70 | 2.59 | 0.95 | 1.25 | .21 | 0.23 | -0.13; 0.58 |
Note. Response scales ranged from 0 (totally disagree) to 5 (totally agree); integrity = specific politician’s integrity; benevolence = specific politician’s benevolence; competence = specific politician’s competence; politicians = trustworthiness of politicians in general; system = trustworthiness of the political system.
Decomposing the interaction effect, we calculated three additional 2 (experimental condition: law violation / control condition) x 2 (integrity vs. benevolence; integrity vs. competence; benevolence vs. competence) ANOVAs with repeated measures on the second factor. Comparing perceived integrity and benevolence, F(1, 419) = 20.01, p < .001, η2p = .05, 95% CI [.01; .09], and comparing integrity and competence, F(1, 419) = 10.84, p < .001, η2p = .03, 95% CI [.00; .06], there were significant interaction effects. When comparing perceived benevolence and competence, the interaction was not significant, F(1, 419) = 1.42, p = .23, η2p = .003, 95% CI [.00; .02]. This pattern of results (see Figure 1) indicates that the effect of the media report about a political law violation was significantly stronger with regard to the politician’s integrity than with regard to his benevolence or competence.
Figure 1
Interaction effect of the experimental condition on the perceived trustworthiness of a specific politician in Study 1.
Evidently, participants used the information about the politician’s law violation to infer character traits (e.g., Clifford, 2014). As suggested by the halo effect (Hypothesis 1c), but in contrast to previous findings (Hendry et al., 2009), the politician’s immoral behavior impaired not only his perceived integrity, but all three trustworthiness dimensions. In line with the predictions from research on trait inferences (e.g., Skowronski & Carlston, 1989), however, the effect on moral integrity was stronger than on benevolence and competence. Notably, the politician’s behavior in our control condition (city councillor approving money for a school renovation) was perceived as legitimate or even prosocial rather than neutral. We will come back to this point in the General Discussion. We tried to overcome this limitation by using a different case in Study 2.
Study 2
In Study 2, we investigated whether a media report of a politician’s law violation had a negative impact on the perceived trustworthiness of politicians in general and the political system.
Method
Sample
Non-psychology undergraduate students (N = 282) were recruited on the campus of a German university for a laboratory study. Data from 14 participants were excluded because they stated that they had not answered the questions honestly or that they had been previously recruited for an identical study. The demographic data of the remaining participants (N = 268) can be found in Table 1. As the two dependent variables perceived trustworthiness of politicians in general and of the political system are theoretically proposed and empirically found to be correlated (e.g., Halmburger, Rothmund, et al., 2019), we used a MANOVA to control for their dependency. Post hoc power analyses of the MANOVA indicated 1 – β = 0.99 for a small (f = 0.10) and β = 1.00 for a medium (f = 0.25) overall effect of the experimental manipulation.
Procedure
The procedure was largely as in Study 1. Participants were randomly assigned to one of two conditions and read a bogus newspaper article that described the behavior of a fictitious German politician. Then, they provided demographic information (age, sex, education, occupation, German language skills). In order to exclude the possibility that our conditions might vary on dimensions other than perceived legitimacy, we add manipulation check items asking for the articles’ credibility and comprehensibility. Participants reported the perceived trustworthiness of politicians in general and of the political system. Finally, they answered two control items regarding their honesty and previous participation at identical studies and were remunerated with five Euros before being thanked and debriefed. All variables can be found in Table 2 of the Electronic Supplementary Material.
Material
All items were answered on 6-point rating scales ranging from 0 (totally disagree) to 5 (totally agree).
Bogus newspaper reports
Participants in the condition with law violation (n = 158) read about how a member of parliament had claimed expenses for private flights in several cases. The district attorney confirmed that the politician was “suspected of criminal activity,” and that he was probing whether this politician illegally charged domestic and international flights to the Hessian state parliament. The politician had lied about the reasons why he spent the money. Participants in the control condition (n = 110) read about how a member of parliament spent his holidays during the summer break. It was described that this politician liked visiting areas he did not know and that he was particularly interested in the varying styles of architecture in different regions (see Electronic Supplementary Material, Figure 3 and 4).
Manipulation checks
We asked participants to indicate how legitimate the politician’s behavior was by using the same four items as in Study 1 (four items, e.g., “I think the politician’s behavior was legitimate; α = .98). In addition, we asked participants whether the article was credible (“The newspaper article was credible”) and comprehensible (“The newspaper article was comprehensible”).
Perceived trustworthiness of politicians in general
We employed a 9-item self-developed scale to assess the three dimensions of trustworthiness, namely integrity (three items, “I believe just now that politicians are sincere”, “At the moment I worry that politicians are not honest”, recoded, “Right now, I doubt that politicians keep their promises” recoded; α = .76), benevolence (three items, “At the moment, I think that politicians attend to the concerns of citizens”, “At the moment I doubt that politicians want the best for the citizens” recoded, “Just now, I worry that politicians do not treat citizens benevolently” recoded; α = .75), and competence (three items, “Just now, I believe that politicians are able to solve political problems”, “At the moment, I doubt that politicians are sufficiently competent”, recoded, “Right now, I worry that politicians are not able to fulfill their political duties”, recoded; α = .61) to assess the perceived trustworthiness of politicians in general. Responses were aggregated across all items to form a measure of perceived trustworthiness of politicians (α = .86).
Perceived trustworthiness of the political system
We used three items to assess the perceived trustworthiness of the political system (“Right now, I believe that justice is promoted in Germany”, “I currently worry about the German democracy”, recoded, “Right now, I doubt that I can trust the German system”, recoded; α = .74).
Control items
Participants were asked if they had answered all questions seriously and honestly (yes/no) and if they had previously been recruited for an identical study (yes/no).
Results and Discussion
Manipulation Checks
Participants who read about a law violation (M = 0.47, SD = 0.71) evaluated the politician’s behavior as significantly less legitimate than those who read the control article without law violation (M = 3.93, SD = 0.90), t(198.52) = 33.66, p < .001, d = 4.36, 95% CI [3.92; 4.80]. In addition, we tested the means against the midpoint of the scale (i.e., 2.5) for both groups separately. Participants in the law violation condition clearly perceived the behavior as illegitimate, t(157) = -35.86, p < .001, d = 2.86. Participants in the control condition violation perceived the behavior as legitimate or prosocial t(109) = 16.66, p < .001, d = 1.59. The articles did not differ in their credibility, t(191.14) = -0.81, p = .42, d = 0.11, 95% CI [-0.14; 0.35], and comprehensibleness, t(266) = -0.05, p = .96, d = 0.01, 95% CI [-0.23; 0.26].
Perceived trustworthiness of politicians in general and the political system
Means, standard deviations, and correlations for all variables are displayed in Table 2. We calculated a MANOVA with the experimental condition as predictor and perceived trustworthiness of politicians in general and the political system as dependent variables. Results showed a significant overall effect of experimental condition, F(2, 265) = 17.74, p < .001, η2p = .12, 95% CI [.05; .19]. Consistent with Hypothesis 2, participants who read about a politician accused of fraud (compared with a politician’s holiday habits) reported significantly lower perceived trustworthiness of politicians in general, F(1, 266) = 35.15, p < .001, η2p = .12, 95% CI [.05; 0.19]. Consistent with Hypothesis 3, participants who read about the law violation (compared with no law violation) also reported lower perceived trustworthiness of the political system, F(1, 266) = 17.13, p < .001, η2p = .06, 95% CI [.02; .12]. The results of the separate t-tests for the dependent measures are displayed in Table 3.
Similar to previous research that investigated how the scandal activation influenced the perception of politicians in general (e.g., Bless et al., 2000; Régner & Le Floch, 2005) or the political system (Anderson & Tverdova, 2003; Bowler & Karp, 2004), our results indicated that a specific politician’s law violation affected more global levels of political trust. These findings can be interpreted in a way that our participants used the negative media reports and assimilated their judgments of the social categories to this information (Bless & Schwarz, 2010). Thus, our findings are in line with the predictions of the inclusion-exclusion model (Bless & Schwarz, 2010). Complementing previous research (e.g., Bless et al., 2000) that tested how the induced salience of a past scandal influenced the perception of politicians in general, Study 2 provided support for direct effects on trustworthiness of politicians in general and the political system by confronting people with new information about a politicians illegal behavior. As in Study 1, our results suggested that the politician’s behavior in our control condition (how a member of the Bundestag spent his summer holidays) was perceived as legitimate or prosocial.
Study 3
Study 3 was designed to replicate the findings of Study 1 (Hypotheses 1a-c) and Study 2, and to test whether the impact on the perceived trustworthiness of politicians in general was mediated by the perceived untrustworthiness of the norm-violating politician (Hypothesis 2).
Method
Sample
A sample that was demographically heterogeneous with regard to their sex, age, and education was recruited by a sampling agency (Keyfacts). Six participants’ data were excluded because they stated that they had searched the internet for more information about the media reports during the ongoing study. The demographic data of the remaining sample (N = 139) are displayed in Table 1. Post hoc power analyses for the conducted ANOVA with manipulation as a first factor, repeated measurement on the second factor, with average correlations of r = .74, and interaction of the between and within-factor, indicated a power of 0.96 for small (f = 0.10) and 1.00 for medium effects (f = 0.25). Given the sample size and the regression coefficient of path a (predictor on mediator) and b (mediator on outcome), mediation analyses with a dichotomous predictor was tested with a power of at least 0.80 (Thoemmes, MacKinnon, & Reiser, 2010).
Procedure
The first part of the study followed the same procedure as the previous studies. Participants read bogus newspaper reports and answered manipulation check items. Then, they indicated how trustworthy they perceived the politician in question and politicians in general to be, among other variables (see Table 3 of the Supplementary Material). Finally, participants answered three control items before being thanked and fully debriefed. All participants were remunerated by the sampling agency.
Material
All items were answered on 6-point rating scales ranging from 0 (totally disagree) to 5 (totally agree).
Bogus newspaper reports
Participants in the condition with law violation (n = 85) read about how a member of the Bundestag had claimed false expenses for private flights and was now accused of fraud in several cases. We used the same experimental stimuli as in Study 2, except the fact that this time the involved politician was a member of the Bundestag (parliament of Germany) instead of a member of the parliament of the federal states. Participants in the control condition (n = 54) read about how a member of the Bundestag showed pupils around the Bundestag building and answered all of their questions. The goal of the trip was for students to experience politics up close and acquire an insight into the occupation of a politician and political processes (see Figure 5 and 6 of the Electronic Supplementary Material).
Manipulation checks
We asked participants to evaluate how illegitimate the politician’s behavior was (four items similar to those in Studies 1 and 2; α = .96). We further asked participants to rate the articles’ credibility (three items; “The article was credible,” “The article was serious,” “The article was real”; α = .90).
Perceived trustworthiness of the politician in question
To enhance the quality of our measurement, we developed a nine-item scale to assess the three dimensions of trustworthiness. Validation studies provided evidence that the proposed three dimensions can be empirically distinguished (Halmburger, Rothmund, et al., 2019). This scales assessed the politician’s perceived integrity (three items, e.g., “Actually, I think that the politician behaved responsibly”; α = .92), benevolence (three items, e.g., “At the moment, I believe that the politician is interested in the opinions of citizens like me”; α = .96), and competence (three items, e.g., “Right now, I am convinced that the politician is well trained”; α = .90). In addition to the three separate dimensions, we aggregated the responses across the dimensions in order to create an overall measure of the perceived trustworthiness of the politician (α = .95).
Perceived trustworthiness of politicians in general
Participants were alerted that the following questions did not refer to a specific politician but to politicians in general. The items used to measure the perceived trustworthiness of politicians in general were identical to those used to assess the perceived trustworthiness of the politician in question (see above), but referred to politicians rather than to a specific politician (nine items; e.g., “Actually, I am convinced that politicians behave responsibly”; α = .95).
Control items
We employed the same control items as in Study 2. Also, participants were asked if they had searched the internet for the media report (yes/no).
Results and Discussion
Manipulation Checks
Participants in the law violation condition evaluated the politician’s behavior as significantly more illegitimate (M = 4.42; SD = 0.71) compared with participants in the control condition (M = 1.27; SD = 1.49), t(68.74) = 14.57, p < .001, d = 2.92, 95% CI [2.43; 3.40]. In addition, we tested the means against the scale’s midpoint (i.e., 2.5) for both groups. Whereas participants in the law violation condition clearly perceived the behavior as illegitimate, t(84) = 24.86, p < .001, d = 2.70, participants in the control condition violation perceived the behavior as rather legitimate or prosocial, t(53) = -6.09, p < .001, d = 0.83. Again, the article with law violation (M = 3.69, SD = 1.04) did not differ from the control article (M = 3.41, SD = 0.96) regarding their credibility, t(137) = 1.57, p = .12, d = 0.28, 95% CI [-0.07; 0.62].
Perceived Trustworthiness of the Politician in Question and Politicians in General
Means, standard deviations, and correlations for all variables are displayed in Table 2. To replicate the findings of Study 1, we tested whether the report on a politician’s fraud affected all dimensions of the respective politician’s perceived trustworthiness compared with the control condition (Hypothesis 1c) by means of a 2 (experimental condition: law violation / control condition) x 3 (dimensions of the perceived trustworthiness of the politician: integrity / benevolence / competence) ANOVA with repeated measures on the latter factor. Mauchly’s test indicated a violation of sphericity for the repeated measures, χ2(2, 136) = 40.77, p < .001. Therefore, we used Greenhouse-Geisser corrected degrees of freedom. There was a significant main effect of the experimental condition, F(1, 137) = 80.37, p < .001, η2p = .37, 95% CI [.25; .47] and of the dimension of perceived trustworthiness, F(1.59, 217.63) = 29.53, p < .001, η2p = .18, 95% CI [.12; .30]. In addition, there was a significant interaction, F(1.59, 217.63) = 12.48, p < .001, η2p = .08, 95% CI [.04; .18].
This pattern of results is in line with Hypothesis 1c, replicating a halo effect found in Study 1, but with a different type of law violation (fraud vs. child pornography) and with the politician in question holding a different political office (member of the Bundestag vs. city councillor).
Decomposing the interaction effect, we calculated three additional 2 (experimental condition: law violation / control condition) x 2 (integrity vs. benevolence; integrity vs. competence; benevolence vs. competence) ANOVAs with repeated measures on the second factor. When comparing integrity and competence, F(1, 137) = 21.99, p < .001, η2p = .14, 95% CI [.05; .25] as well as perceived benevolence and competence F(1, 137) = 8.86, p = .003, η2p = .06, 95% CI [.01; .15], there were significant interaction effects. In contrast, for the perceived integrity and benevolence, F(1, 137) = 3.14, p = .079, η2p = .02, 95% CI [.00; .09], no interaction effect was found. This pattern of results indicates that the effect of the media report about a political law violation was significantly stronger with regard to the politician’s integrity and benevolence than with regard to his competence.
In order to test whether the report of the political law violation negatively affected the perceived trustworthiness of politicians in general, we calculated an independent samples t-test with the perceived trustworthiness of politicians as the dependent variable (Table 3). In contrast to the findings from Study 2, the main effect of the law violation on the perceived trustworthiness of politicians was not significant, t(137) = 1.46, p = .15, d = 0.25, 95% CI [-0.09; 0.56]. This is particularly surprising, since the article in the condition without law violation (member of the Bundestag showed pupils around the Bundestag) is still perceived as legitimate or prosocial rather than neutral. Based on this finding, the direct effect of the experimental condition should be even more pronounced than when comparing an illegal with a neutral behavior of a politician.
Indirect Effect
Despite the non-significant total effect of the experimental condition on the perceived trustworthiness of politicians in general, we wanted to test the hypothesized psychological process of assimilation in the context of political law violations. Therefore, we investigated whether the law violation had an indirect effect on the perceived trustworthiness of politicians in general via the aggregated perceived trustworthiness of the politician in question (Hypothesis 2). We used the process macro as a bootstrapping procedure with 1,000 resamples (Hayes, 2012). Statistical significance was determined by inspecting the respective 95% confidence intervals (Preacher & Hayes, 2004). Again, the total effect of the experimental condition (0 = control condition, 1 = law violation) on the perceived trustworthiness of the politician in general was not significant, B = -0.28, SE = 0.19, p = .14, 95% CI [-0.68; 0.09], but there was a negative indirect effect of the experimental condition on the perceived trustworthiness of politicians in general via the perceived trustworthiness of the politician in question (Figure 2), B = -0.80, SE = 0.16, 95% CI [-0.54; -1.19]. Participants in the law violation condition perceived the politician in question to be significantly less trustworthy compared with participants in the control condition, B = -1.51, SE = 0.17, p < .001. The perceived trustworthiness of this individual politician was positively associated with the perceived trustworthiness of politicians in general, B = 0.53, SE = 0.09, p < .001. Thus, we found evidence in partial support of the idea that the decrease of perceived trustworthiness of politician in general is shaped by the decrease of trustworthiness by the politician in question.
Figure 2
Indirect effect of the experimental condition on the perceived trustworthiness of politicians in general mediated by the perceived trustworthiness of the involved politician in Study 3.
Note. N = 139. The figure displays unstandardized regression coefficients. The number in parentheses represents the coefficient for the total effect.
*p < .05. ***p < .001.
Strikingly, however, there was an additional positive direct effect of the experimental condition on the perceived trustworthiness of politicians in general, B = 0.52, SE = 0.22, p = .02, when controlling for the indirect effect. Statistically, this can be interpreted as a suppression effect due to the shared variance of the perceived trustworthiness of the specific politician and politicians in general. In line with the inclusion-exclusion model (Bless & Schwarz, 2010), this result pattern indicates that contrast effects have accompanied the indirect effects. We will discuss this result in more detail in the General Discussion. We also tested whether the mediation effect worked in the opposite direction, finding no significant indirect effect of the experimental condition on the perceived trustworthiness of the politician in question via the perceived trustworthiness of politicians in general, B = -0.11, SE = 0.08, 95% CI [-0.33, 0.03].
Study 4
In Study 4, we again tested Hypotheses 1a-c to replicate the findings from Studies 1, 2, and 3. Moreover, we tested whether reading about a political law violation affected the perceived trustworthiness of the political system and whether this effect was mediated by the perceived trustworthiness of the politician in question (Hypothesis 3).
Method
Sample and Procedure
N = 122 students were recruited to participate in a laboratory experiment via the university’s mailing lists (see Table 1 for demographic data). For the conducted ANOVA with manipulation as a first factor, repeated measurement on the second factor with average correlations of r = .64 for the repeated measures, and interaction of the between and within-factor, post-hoc power analyses indicated a power of 0.82 for small (f = 0.10) and 1.00 for medium effects (f = 0.25). Given the sample size and the regression coefficients, mediation analyses with a dichotomous predictor was tested with a power of at least 0.80 (Thoemmes et al., 2010). The first part of the study followed the same procedure as the previous studies. After participants had read the bogus newspaper report and answered the manipulation check items, they indicated the perceived trustworthiness of the politician in question and of the political system. Finally, participants answered one control item, were remunerated with five Euros, and fully debriefed. All variables can be found in Table 4 of the Electronic Supplementary Material.
Material
If not reported otherwise, all items were answered on 6-point rating scales ranging from 0 (totally disagree) to 5 (totally agree).
Bogus newspaper reports
For the experimental manipulation, similar to Study 1, participants in the law violation condition (n = 64) read about how a city councillor was accused of having possessed and exchanged child pornography in several cases. The participants in the control condition (n = 58) read about how a city councillor approved money for a school renovation (see Figure 7 and 8 of the Electronic Supplementary Material).
Manipulation check
As in the previous studies, we asked participants to rate how legitimate the politician’s behavior was (four items; α = .98).
Perceived trustworthiness of the politician in question
We used the same 9-item scale as in Study 3, which included items for the three dimensions of trustworthiness (α = .87 - .94). Cronbach’s alpha for the aggregated scale was .93.
Perceived trustworthiness of the political system
We then used a self-developed 9-item scale for the perceived trustworthiness of the political system. In line with the theoretical conceptualization of trustworthiness (Mayer et al., 1995; Mayer & Davis, 1999; Schoorman et al., 2007), we distinguished between three dimensions of trustworthiness, namely, fairness of the system (three items, e.g., “At the moment, I trust that all people are treated equally in the German political system”; α = .88), responsiveness of the system (three items, e.g., “Right now, I assume that all citizens are able to participate in important decisions in our political system”; α = .81), and effectiveness/efficacy of the system (three items, e.g., “At the moment, I believe the German political system is efficient”; α = .83). Cronbach’s α for the whole scale was .88.
Results and Discussion
Manipulation Check
Participants in the law violation condition evaluated the politician’s behavior as less legitimate (M = 0.56; SD = 0.91), compared with participants in the control condition (M = 3.78; SD = 0.99), t(120) = 18.80, p < .001, d = 3.39, 95% CI [2.34; 3.95]. In addition, we tested the means against the scale’s midpoint (i.e., 2.5) for both groups. Whereas participants in the law violation condition clearly perceived the behavior as illegitimate, t(63) = -17.14, p < .001, d = 2.13, participants in the control condition perceived the behavior as legitimate or prosocial t(57) = 9.88, p < .001, d = 1.29.
Perceived Trustworthiness of the Politician in Question and the Political System
Means, standard deviations, and correlations for all variables are displayed in Table 2. Regarding the perceived trustworthiness of the politician in question, we calculated a similar ANOVA as in the previous studies. There was a significant main effect of our experimental manipulation, F(1, 120) = 115.31, p < .001, η2p = .49, 95% CI [.36; .58]. Consistent with Hypothesis 1c, participants in the law violation condition perceived the politician in question to be significantly less trustworthy on each dimension compared with those in the control condition (see Table 3). Furthermore, there was a significant main effect of the dimension of perceived trustworthiness, F(2, 119) = 5.95, p = .003, η2p = .09, 95% CI [.01; .19] and a significant interaction between the experimental conditions and the dimensions of trustworthiness, F(2, 119) = 17.08, p < .001, η2p = .22, 95% CI [.10; .34].
Decomposing the interaction effect, we calculated three additional 2 (experimental condition) x 2 (integrity vs. benevolence; integrity vs. competence; benevolence vs. competence) ANOVAs with repeated measures on the second factor. Comparing perceived integrity and benevolence, F(1, 120) = 12.66, p < .001, η2p = .10, 95% CI [.02; .20], and integrity and competence, F(1, 120) = 29.43, p < .001, η2p = .20, 95% CI [.08; .32], there was a significant interaction between the experimental conditions and the dimensions of perceived trustworthiness. When comparing benevolence and competence, the interaction was not significant, F(1, 120) = 0.86, p = .34, η2p = .01, 95% CI [.00; .06]. Thus, as in Study 1, the effect of the media report about a political law violation was stronger with respect to the integrity of the politician in question than regarding his benevolence and competence.
Regarding the perceived trustworthiness of the political system, we calculated an independent samples t-test (see Table 3). Unlike in Study 2, the effect of the law violation was not significant, t(115.40) = 1.27, p = .21, d = 0.23, 95% CI [-0.13; 0.58]. We will discuss this inconsistency in more detail in the General Discussion.
Indirect Effect
In order to test indirect effects of the experimental conditions on the perceived trustworthiness of the political system via the perceived trustworthiness of the specific politician, the aggregated scales were used. Although the total effect of the experimental condition (0 = control condition, 1 = law violation condition) on the perceived trustworthiness of the political system was not significant, B = -0.19, SE = 0.15, p = .21, 95% CI [-0.48; 0.97], our results revealed a significant indirect effect of the experimental condition on the perceived trustworthiness of the political system via the perceived trustworthiness of the politician in question (Figure 3), B = -0.58, SE = 0.17, 95% CI [-0.26, -0.90]. Besides the indirect effect, there was a non-significant positive direct effect of the experimental condition on the perceived trustworthiness of the political system, B = 0.39, SE = 0.20, p = .06. This pattern of results is contradictory to what we would have assumed on the basis of the inclusion-exclusion model (Bless & Schwarz, 2010).
Figure 3
Indirect effect of the experimental condition on the perceived trustworthiness of the political system mediated by the perceived trustworthiness of the involved politician in Study 4.
Note. N = 122. The figure displays unstandardized regression coefficients. The number in parentheses represents the coefficient for the total effect.
***p < .001.
We also tested whether the mediation worked in the opposite direction and found no significant indirect effect of the experimental condition on the perceived trustworthiness of the politician in question via the perceived trustworthiness of the political system, B = -0.06, SE = 0.06, 95% CI [-.20, .03].
General Discussion
An increasing negativity bias in media reports on politicians has been blamed for a decline in political trust in the general public (Cappella & Jamieson, 1997; Lengauer et al., 2011; Moy & Pfau, 2000). The present research tested implications of this assumption in four experimental studies. We investigated whether and how media reports about a politician’s law violation affected the perceived trustworthiness of the politician in question, of politicians in general, and of the political system.
Indicating a halo effect (e.g., Thorndike, 1920), we found that reports about a politician’s violation of law reduced not only the politician’s perceived integrity, but also his perceived benevolence and competence. More specifically, this effect can be understood as a true halo effect (Murphy et al., 1993), as the affected dimensions are not unrelated, but build a meaningful concept of trustworthiness (see also Halmburger, Rothmund, et al., 2019). This finding was consistent across three studies involving different domains of law violations and different measures of perceived trustworthiness. Our studies provided compelling evidence that people use information about the law-violating behavior of individual politicians in the media as a source for evaluating those politicians’ general trustworthiness. As all three dimensions have been found to be important and independent predictors of citizens’ voting behavior (Funk, 1999; Halmburger, Rothmund, et al., 2019; Hayes, 2010; Mölders et al., 2017), our research emphasizes the severity of the damage a politician has to fear when his or her unlawful behavior is uncovered, particularly, shortly before an election (Mitchell, 2014). Two out of three studies further suggested that the politician’s perceived integrity was impaired more than his competence and in two studies the politician’s perceived integrity was also impaired more than his benevolence (see also Hendry et al., 2009). Thus, in line with research on trait inferences (Skowronski & Carlston, 1989; Wojciszke, Bazinska, & Jaworski, 1998), it can be argued that a politician’s violation of law might be particularly informative for this person’s perceived integrity.
As a second research question, we investigated whether and how information about a single politician’s law violation could impair the perceived trustworthiness of politicians in general and the political system. In Study 2, we tested direct effects of a single politician’s law violation on the perceived trustworthiness of politicians in general and of the political system. Study 2 supported our expectations by showing that the perceived trustworthiness of politicians in general and of the political system were lowered after recipients had read about a politician’s violation of law. This is in line with previous research suggesting that information about a politician’s wrongful behavior is generalized to the untrustworthiness of politicians in general and of the political system (e.g., Halmburger et al., 2012). In Studies 3 and 4, we tested whether this relation is mediated by the perceived untrustworthiness of the politician that violated the law. As hypothesized, we found indirect effects of the experimental conditions on the perceived trustworthiness of politicians in general (Study 3) and on the perceived trustworthiness of the political system (Study 4) via a decrease in the perceived trustworthiness of the politician involved in the violation of law. These results were inconsistent with the possibility that the process could work in the opposite direction, because the perceived trustworthiness of politicians in general and the political system, respectively, did not mediate effects on the perceived trustworthiness of the politician in question. This pattern of results suggests that the indirect effects that we found were not due merely to a conceptual overlap between the trustworthiness measures, but instead reflected a meaningful psychological process.
Despite the fact that the indirect effects observed in Studies 3 and 4 were in line with our prediction, it is important to note the non-significant total effects of the experimental manipulation in these studies. In Study 3, we even found a positive direct effect of law violation on the perceived trustworthiness of all politicians, when controlling for the indirect effect via the perceived trustworthiness of the involved politician. In total, however, the effect of the law violation on perceived trustworthiness of all politicians was insignificant. Statistically, this pattern of results could suggest a suppression effect. The occurrence of such an effect is likely due to strong conceptual overlap between the mediator (perceived trustworthiness of the specific politician) and the dependent variable (perceived trustworthiness of all politicians). In line with the inclusion-exclusion model (Bless & Schwarz, 2010), this finding can be interpreted in a way that our participants excluded the information about the specific politician from the evaluation of all politicians in Study 3. In contrast to this politician, people perceived all politicians as even more trustworthy. Considering the boundary conditions of assimilation and contrast effects, we can assume that participants perceived the information in this study to be less relevant and representative for the social category. In line with the empirical findings of Herr, Sherman, and Fazio (1983), conversation rules might have caused a contrast effect. After being asked about the trustworthiness of the law-violating politician, participants might have actively corrected their judgments about politicians in general, and might have done so to different degrees.
In Study 4, we found neither a negative nor a positive significant direct effect of the report on the perceived trustworthiness of the political system. Similar to Study 2, this pattern suggests a suppressor effect. This finding can be interpreted in a way that no assimilation or contrast effects occurred or that both effects occurred at the same time and cancelled each other out. In line with the exclusion-inclusion model (Bless & Schwarz, 2010), this result suggests that participants did not believe that the information about the specific politician was meaningfully related to the political system and therefore had no impact.
Taken together, a negative effect of media reports on a specific politician on politicians in general, and the political system could only be found when we did not assess the trustworthiness of the involved politician. As an alternative explanation for our results and previous ones (e.g., Schwarz & Bless, 1992), it can be discussed whether the effect on the social category is rather artificial. People seem to be motivated to express their negative evaluation of the specific politician’s behavior. They could even have experienced demand effects when asked to evaluate a specific politician, politicians in general, or the political system directly after reading about an unlawful behavior. If they do not have the opportunity to evaluate the responsible person directly, they have no other means but to evaluate the superior category. As soon as people have the opportunity to evaluate the person responsible, the assimilation effect disappears. In this case, the target information could be perceived as irrelevant for the social category and, in consequence, the evaluation of the social category should be unaffected or even more positive (if contrast effects occur). It is important to note, however, that in all our studies, participants were first asked to evaluate the politician’s behavior as moral or immoral on manipulation check items. So, on the other hand, counteracting mechanisms might only be activated when very similar questions are asked regarding the trustworthiness of a concrete politician and of all politicians (or of the system) in general. Clearly, more research would be helpful to shed light on the exact nature and boundary conditions of mechanisms counteracting such a detrimental effect of media reports of politicians’ wrongdoings.
Limitations and Future Research
When drawing conclusions from our findings, some potential limitations need to be kept in mind. First, similar to previous research (Funk, 1996; Mitchell, 2014; Mölders et al., 2017), the current studies involved fictitious newspaper reports about unknown politicians. To secure the generalizability of our results, we employed reports that were very similar to past reports about real political scandals, and we varied the type of law violation and the office of the politicians in question across our studies. The results of our manipulation checks suggested that our participants perceived these articles as credible. Whereas the experimental design clearly represents the strength of our studies, we are aware that it comes with limitations for the external validity. As an important limitation, it needs to be noted that our material exclusively focused on the impact of the illegal behavior of male politicians. Whether the pattern of results would generalize to female politicians still needs to be scrutinized. Due to the impact of stereotypic roles, it is possible that illegal, immoral, or incompetent behaviors of politicians might be evaluated differently in dependence of the politician’s gender (Huddy & Terkildsen, 1993). We suggest that future studies should aim to replicate our findings with reports about the behavior of real politicians (male and female), while making use of our precise measures of trustworthiness.
In our studies, the experimental stimuli used for the control condition (i.e., approving money for a school renovation) were perceived as legitimate or prosocial (rather than neutral) behavior on the part of the politician, as indicated by the manipulation checks in all four studies. On the one hand, this could have enhanced differences between the experimental conditions in the perceived trustworthiness of the politicians involved. Potentially, the observed effects of the experimental conditions could have been driven by characteristics of the control condition. However, the politician’s behavior in the norm violation condition was rated as much more illegitimate (as judged in terms of deviation from the midpoint of the scale) compared to how legitimate the politician’s behavior in the control condition was judged. Thus, even if the slightly positive perception of the control condition enlarged effects on perceived trustworthiness, it seems unlikely that our results could be only or predominantly attributed to characteristics of the control condition. On the other hand, the pattern of results on the manipulation checks makes the null findings on the perceived trustworthiness of politicians in general (Study 3) and the political system (Study 4) even more relevant. In our research, this perception of the control condition as prosocial was unintended.
To the best of our knowledge, the impact of positive media reports on political trustworthiness is under-researched. It would be highly interesting to investigate the impact on political trustworthiness by systematically manipulating the valence of media information. Our empirical approach could be extended by testing systematically the boundary conditions of the effects that we identified. We found robust effects on the perceived trustworthiness of the politicians in question across studies that involved different kinds of law violations (financial fraud vs. child pornography). However, we did not compare these kinds of violations in a single study (cf., Carlson et al., 2000; Funk, 1996).
Furthermore, future studies could investigate the impact of a politician’s mistakes or unresponsive behaviors, which might be particularly damaging to the politician’s perceived competence or benevolence, respectively. Empirical findings (Funk, 1996) suggest that the impact of information on a lack of competence compared with a lack of integrity on the perception of a politician’s character could be even worse. However, up to now, there are too few studies directly comparing different types of negative information on different dimensions of trustworthiness to draw final conclusions. As already discussed, the impact of positive media information about a specific politician would be highly interesting.
Our participants received negative information about an unknown politician whose party affiliation was not mentioned. By this, we aimed at controlling for previous knowledge, attitudes, and party identification. In line with previous findings (Funk, 1996), however, it can be assumed that a positive attitude towards a specific politician’s trustworthiness can work as a buffer against the negative influence of media reports. Such interaction patterns can only be investigated by assessing changes of the perceived trustworthiness in longitudinal designs. In addition, further contextual information such as the representativeness of a specific politician for all politicians or the political system has been found to influence information processing (Strack, Schwarz, & Gschneidinger, 1985; see Schwarz & Bless, 1992). In line with the exclusion-inclusion model (Bless & Schwarz, 2010), it can be argued that assimilation processes should increase when politicians are perceived as highly representative for the social category. Importantly, our studies showed inconsistent results with regard to direct effects of media reports about a specific politician on the trustworthiness of all politicians or the political system. From this point of view, it would be interesting to gain insights into the psychological mechanisms that restrict the impact of media reports for the perception of the social group or system. Although we speculated about the causes of these inconsistencies, they need to be investigated empirically in future studies. One possible idea would be to assess all levels of trustworthiness in one study, but varying the order of questions (Schwarz & Bless, 1992).
Our research is also limited by the fact that we focused on immediate reactions to media reports of one politician’s immoral behavior. Research in social psychology has suggested that the repeated confrontation with untrustworthiness promotes the development of generalized distrust expectations (Rothmund, Gollwitzer, Bender, & Klimmt, 2015). Although the inclusion-exclusion model (Bless & Schwarz, 2010) does explicitly capture such cases, it seems plausible to expect that the repeated confrontation with negative media reports about specific politicians shape the way we perceive politicians in general and the political system. Nevertheless, future studies are needed to directly test the extent to which repeatedly inserting negative information (e.g., norm violation, law violation) into a mental representation of a broader category (e.g., trustworthiness of politicians in general) will lead to an enduring change in this mental representation.
Conclusions
Taken together, media reports on the wrongdoings of a specific politician have the potential to reduce citizens’ expectations that this politician will act or function in a morally right, competent, and benevolent manner. How much people generalize from the wrongdoing of one politician to all politicians or the political system seems to be highly influenced by boundary conditions.